Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

Vattila · Oct 6, 2019

Except for the details about the improvements in the microarchitecture, we now know pretty well what to expect with Zen 3.

The leaked presentation by AMD Senior Manager Martin Hilgeman shows that EPYC 3 "Milan" will, as promised and expected, reuse the current platform (SP3), and the system architecture and packaging looks to be the same, with the same 9-die chiplet design and the same maximum core and thread-count (no SMT-4, contrary to rumour). The biggest change revealed so far is the enlargement of the compute complex from 4 cores to 8 cores, all sharing a larger L3 cache ("32+ MB", likely to double to 64 MB, I think).

Hilgeman's slides did also show that EPYC 4 "Genoa" is in the definition phase (or was at the time of the presentation in September, at least), and will come with a new platform (SP5), with new memory support (likely DDR5).

What else do you think we will see with Zen 4? PCI-Express 5 support? Increased core-count? 4-way SMT? New packaging (interposer, 2.5D, 3D)? Integrated memory on package (HBM)?

Vote in the poll and share your thoughts!

RnR_au · May 23, 2022

gdansk said:
In the presentation it sounds like 1T uplift is ">15%" including the clock speed increase.
Well that's less than expected.

She used the words 'significantly greater than 5Ghz', which kinda doesn't match the slide.

Saylick · May 23, 2022

She's sandbagging. She's setting up a surprise for the real desktop Zen 4 launch later this year.

"Significantly higher clocks than 5 GHz"
"Higher IPC"
">15% ST"

The clocks alone should be almost 10% uplift. I find it incredibly hard to believe that the IPC gains are less than double digit.

Edit: 16C Zen 4 shown off running Ghostwire: Tokyo at 5.52 GHz. Also shown off running Blender. 16 core Zen 4 beats the 12900K 31% faster.

Tuna-Fish · May 23, 2022

5.52GHz shown during gameplay.

(edit: ) Well that had almost zero detail. It was really more of a teaser of the AM5 platform.

Saylick · May 23, 2022

The footnotes on that >15% ST claim says it's a 16C Zen 4 against the 5950X in Cinebench R23 1T.

gdansk · May 23, 2022

Saylick said:
The footnotes on that >15% ST claim says it's a 16C Zen 4 against the 5950X in Cinebench R23 1T.

It doesn't say the score? If it's merely 1.16 the 5950X it's not ahead of the 12900K

DisEnchantment · May 23, 2022

Lel, 5% IPC gain.

Saylick · May 23, 2022

gdansk said:
It doesn't say the score? If it's merely 1.16 the 5950X it's barely ahead of the 12900K

Nope. No score. Just listed the benchmark and the system that was used for the claim.
I would put emphasis on the ">" symbol. Not that I know anything from an insider but there's got to be a reason it's there. The clocks already tell you half the picture.

tamz_msc · May 23, 2022

Saylick said:
The footnotes on that >15% ST claim says it's a 16C Zen 4 against the 5950X in Cinebench R23 1T.

That's still slower in Cinebench than 12900K. But 5950X is already faster than 12900K in Blender. It must mean that Cinebench, which was a big improvement with Alder Lake, isn't as big an improvement on Zen 4.

Saylick · May 23, 2022

Here's the footnote for reference. My apologies if it's a little hard to read.

gdansk · May 23, 2022

Also no room for a 3rd CCD but that's not surprising.

tamz_msc · May 23, 2022

If general IPC increase is indeed small, then all the xtor budget must have been spent on increasing clock speeds.

Vope45 · May 23, 2022

tamz_msc said:
That's still slower in Cinebench than 12900K. But 5950X is already faster than 12900K in Blender. It must mean that Cinebench, which was a big improvement with Alder Lake, isn't as big an improvement on Zen 4.

I find it very weird

In this video 12900k is about 9% faster than 5950x

Also this too: https://www.techpowerup.com/review/intel-core-i9-12900k-alder-lake-12th-gen/6.html

12900k is about 7% faster than 5950x in blender

I find it particular hard to believe with Ipc increase is only at 5%

coercitiv · May 23, 2022

Saylick said:
I would put emphasis on the ">" symbol. Not that I know anything from an insider but there's got to be a reason it's there. The clocks already tell you half the picture.

We can put emphasis any way we like, it still looks weak. Then again I see quite a number of people gloating and picturing Raptor Cove will stump all over Zen4 while they forget Intel themselves made similarly vague performance claims to what AMD does today:

Personally I gave AMD the benefit of the doubt until I saw the presentation today. I don't like what I'm seeing.

tamz_msc said:
If general IPC increase is indeed small, then all the xtor budget must have been spent on increasing clock speeds.

If I see another speed demon design I'm gonna blow a fuse.

lobz · May 23, 2022

Tuna-Fish said:
I think the big thing is AVX512. When not using that, the uplift will be fairly small, maybe that 15%.

When using AVX512, the uplift will be much more substantial.

wrong

Saylick · May 23, 2022

Some really awful napkin math here, but just trying to see what kind of IPC gains are required to get 31% more performance in Blender than the 12900K.

If we assume that Gracemont has an IPC value of 1 or roughly similar to Skylake, Golden Cove should be 1.4 or so. If we assume that the clocks for the 12900K are 5 GHz for GC and 4 GHz for GM, that's a "score" of: 8x5x1.4 + 8x4x1 = 88.

Meanwhile, if we assume that Zen 4 has roughly the same IPC as Golden Cove and runs at 5 GHz all core, the "score" is: 16x5x1.4 = 112.

The ratio is: 112/88 = 1.27.

Again, super rough math and some liberal assumptions, but it's clear that the IPC uplift has to be in the double digits. The >15% ST claim is just that. Greater than 15%.

coercitiv · May 23, 2022

Saylick said:
Some really awful napkin math here, but just trying to see what kind of IPC gains are required to get 31% more performance in Blender than the 12900K.

You don't need the napkin math though, we have benchmarks for both 12900K and 5950X in Blender. The 12900K is ~5% faster than 5950X in the Techpowerup test and about equal in the Gamer Nexus test.

Vope45 · May 23, 2022

coercitiv said:
You don't need the napkin math though, we have benchmarks for both 12900K and 5950X in Blender. The 12900K is ~5% faster than 5950X in the Techpowerup test and about equal in the Gamer Nexus test.

View attachment 61939

View attachment 61940

Unless amd test it with e core disabled ?!!!!

coercitiv · May 23, 2022

Vope45 said:
Unless amd test it with e core disabled ?!!!!

And a nail driven through the socket. Let's stay within the realm of reason, AMD isn't Principled Technologies.

poke01 · May 23, 2022

Saylick said:
Some really awful napkin math here, but just trying to see what kind of IPC gains are required to get 31% more performance in Blender than the 12900K.

If we assume that Gracemont has an IPC value of 1 or roughly similar to Skylake, Golden Cove should be 1.4 or so. If we assume that the clocks for the 12900K are 5 GHz for GC and 4 GHz for GM, that's a "score" of: 8x5x1.4 + 8x4x1 = 88.

Meanwhile, if we assume that Zen 4 has roughly the same IPC as Golden Cove and runs at 5 GHz all core, the "score" is: 16x5x1.4 = 112.

The ratio is: 112/88 = 1.27.

Again, super rough math and some liberal assumptions, but it's clear that the IPC uplift has to be in the double digits. The >15% ST claim is just that. Greater than 15%.

Please don't look at first party benchmarks. Always look for third party benchmarks.

Blender is where Intel was on par or ahead of AMD using 3rd party benchmarks

We can see from AMD themselves that on Cinebench R23 its around 15% uplift in ST when compared to 5950X.

That is disappointing. 2 years, a new node and all the budget has been spent on increasing clocks. Zen 3 to Zen 4 is not a huge arch improvement like most had hoped.

It may have AVX 512 useful for the 0.1% of PC users.

Karnak · May 23, 2022

Vope45 said:
Unless amd test it with e core disabled ?!!!!

Su said 16C for both. So I doubt that E cores were disabled.

deasd · May 23, 2022

It's either Cinebench ST is not the goal that AMD aimed for, or the SMT performance with Zen4 is larger than expected which would be enough to generate 31% advantage against 12900k in Blender.

biostud · May 23, 2022

coercitiv · May 23, 2022

deasd said:
It's either Cinebench ST is not the goal that AMD aimed for, or the SMT yield with Zen4 is larger than expected which would be enough to generate 31% advantage against 12900k in Blender.

Blender is not a good indicator unless used with the same test scene, since performance varies greatly from one scene to another. While Techpowerup and GN found 5950X to be equal or a bit faster than 12900K in their respective tests, Hardware Unboxed found 59050X to be ~20% faster than 12900K using another Blender test file.

With all it's faults Cinbench is still a more reliable performance indicator than non-standardized Blender testing.

JoeRambo · May 23, 2022

So Gigabyte leak was spot on. 1MB of L2 cache (with same 8-ways sadly), minor updates to core ( increased L2 TLB ) and AVX512 support.

I've always wondered where some guys here are pulling those outrageous claims of "30% IPC increase and 5.5Ghz clocks". The end result is what was predicted by the leaks: Zen3 with 1LMB of L2, same 32MB of L3 on release, DDR5 and 5-10% IPC increase.

Timmah! · May 23, 2022

NostaSeronx said:
Zen4 with V-Cache is end of era.

Zen3 w/ V-cache => Last Zen3 CPU before Zen4
Zen4 w/ V-cache => Last Zen4 CPU before Zen5

If anyone sees V-cache it means the next-generation is around the corner.

For example, 4th Gen EPYC with V-Cache hasn't been officially announced with a codename yet.

well, i thought that 5800x3d was sort of a test run for the v-cache technology and it would become standard by 7000 generation, at least on say ryzen 9 parts.

Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

Senior member

Golden Member

Diamond Member

Golden Member

Diamond Member

Platinum Member

Golden Member

Diamond Member

Diamond Member

Diamond Member

Platinum Member

Diamond Member

Member

Diamond Member

Platinum Member

Diamond Member

Diamond Member

Member

Diamond Member

Senior member

Senior member

Senior member

Lifer

Diamond Member

Golden Member

Golden Member